Late-Stage Artificial Problem?

Technological revolutions tend to move through three stages. First comes introduction: an innovation is born and tested in relative obscurity. Then permeation: it spreads into mainstream use across industries. Finally, power: it becomes so embedded in daily life that removing it would be unthinkable. Electricity followed this arc between roughly 1880 and 1920, alongside the telephone, the railroad, the automobile, and cinema. Farms emptied into cities, global trade upended local livelihoods, and fortunes were made and lost within a generation.

We are living through our own revolution, and it has followed the same script, just written in code rather than copper wire. Computers were the introduction. The internet and smartphones drove permeation. Artificial intelligence is now circling the power stage, the point where a technology stops being merely exciting and starts becoming essential infrastructure. Demis Hassabis, CEO of Google DeepMind, described it: we’ve found a way to make sand think, since chips are, after all, made of sand. Could we be witnessing the dawn of that power stage, or the anxious final act of a revolution.

Investment markets have been asking the same question. A small number of AI and semiconductor companies has captured both the excitement and anxiety of the revolution, delivering extraordinary returns alongside extraordinary swings.

In July 2026 that anxiety found its evidence. Moonshot, a Chinese startup, released a high performing, freely available AI model that rivalled the best coming out of Silicon Valley. The Philadelphia Semiconductor Index, home to Nvidia, Broadcom, and Micron, fell roughly 10% in a single week, its sharpest drop since April 2025. It wasn’t just the model itself. It was the fear that cheap, capable foreign AI could erode assumed dominance and invite tighter export controls, unsettling supply chains that investors had priced in.

What actually moved the market wasn’t a profit warning or an interest rate decision. It was doubt about whether a lead everyone had assumed was secure might be more fragile than it looked. In a market this concentrated, that kind of doubt spreads fast, because so few companies are propping up so much of the index.

The Magnificent Seven (who have underpinned the revolution) now account for roughly 40% of the S&P 500, meaning the ten largest holdings outweigh the bottom 490 combined. Nvidia trades near 50 times earnings, Tesla above 240, and Apple around 37 despite modest growth. The Shiller CAPE ratio cleared 40 in 2025, a level last seen just before the dot com crash.

Is this 2000 again? Not quite. Many Dot Com companies were unprofitable, running on promises real earnings. Today’s AI leaders actually make money. Nvidia alone posted close to $99 billion in profit last year, and the big cloud computing companies are funding their AI buildouts from their own cash flow rather than borrowed money. That’s evidence the last bubble lacked.

But does the math add up? Roughly $400 billion in 2025 AI capital expenditure against only about $100 billion in AI revenue, a four to one gap. Are markets exposed? What if another rival model comes along?

None of this is an argument against AI. It’s an argument for knowing how exposed you already are, since that exposure often hides inside an index fund or thematic strategy that looks ‘diversified’ while quietly leaning on the same few names. The real question isn’t whether AI succeeds. It’s whether your portfolio can weather the concentration risk. Diversification won’t win every year, but it keeps your future protected. Revolutions, after all, have a habit of arriving with fireworks and eventually settle into just another chapter in the history books.

If you’re unsure how much of your portfolio is riding on a handful of AI stocks, let’s talk. Reach out to schedule a conversation with an adviser.